Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Back to Podcast Digest
Did Apple Get AI Spending Right?, Microsoft & OpenAI’s New Reality, Where’s Stargate?
TL;DR
Apple’s tiny AI capex is either discipline or denial — MG Seigler notes Apple may spend just $9–10 billion in capex this year versus roughly $145 billion at Meta and around $190 billion guidance from Amazon and Google, making Apple the clear outlier in Big Tech’s AI arms race.
John Ternus is already signaling Apple may loosen the purse strings — Alex and MG read Apple’s new flexibility around buybacks, dividends, and cash neutrality as an early clue that Ternus could fund bigger AI infrastructure, R&D, or acquisitions once he takes over.
Apple’s real bet is that hardware still wins even if it doesn’t own the best model — The core bullish case is that the iPhone, Macs, and Apple silicon remain the main place people actually use AI, especially if on-device models become “good enough” before standalone AI devices truly matter.
Microsoft and OpenAI’s latest deal looks like a controlled separation, not a reunion — By killing the AGI clause and letting OpenAI sell through AWS—and likely Google Cloud next—Microsoft appears to be trading exclusivity for a cleaner 2032 IP arrangement plus upside from its huge equity stake.
Microsoft gave up a lot, but not everything — MG highlights the fine print that OpenAI products still ship first on Azure unless Microsoft can’t or won’t support them, meaning Azure may keep a right-of-first-refusal advantage even as OpenAI goes multicloud.
Stargate has quietly shifted from a giant OpenAI-owned buildout to a branding layer over partner infrastructure — The original $500 billion, White House-backed vision with Trump, Sam Altman, Masayoshi Son, and Larry Ellison has morphed into OpenAI securing access through Oracle, Nvidia, SoftBank, and neocloud partners rather than clearly building and owning the data centers itself.
The Breakdown
Apple’s weirdly low AI spend finally looks intentional
Alex opens with the striking contrast in Big Tech capex: Apple is heading toward roughly $9–10 billion this year while Meta is around $145 billion and Amazon and Google may each approach $190 billion. MG says the chart tells the whole story — everyone’s spending exploded after the ChatGPT era, while Apple’s line just sits there, almost flat, like it missed the memo.
The “binary bet”: Apple could look brilliant or deeply exposed
MG frames Apple’s position as a real binary bet. Maybe Apple is wisely waiting for model costs to commoditize, leaning on partners or open source the way it benefited from Google Search; or maybe it’s setting itself up to become a legacy device company begging OpenAI or Google for the tech that powers the next interface.
Why Apple may be admitting it can’t win the frontier-model race
Alex argues the market has learned that “spend $100 billion and get a great model” just isn’t true — Amazon Nova barely registers, Microsoft hasn’t replaced OpenAI, Meta is still searching, and xAI hasn’t broken through. MG agrees Apple may simply be making the sober call that it lacks the talent and timing to catch up, so why burn tens or hundreds of billions trying.
Apple’s best AI argument is still the iPhone in your hand
Despite all the hand-wringing, MG still thinks Apple could be an AI winner because the iPhone remains the default computing hub and Apple makes the best hardware. He and Alex both land on the same practical point: the best AI device is often just the one already in your hand, not the hypothetical Jony Ive gadget that might replace it someday.
Ternus starts showing his hand
The conversation then turns to John Ternus, with Alex pointing to his early AI-heavy messaging and Apple’s new financial optionality as a meaningful signal. MG says loosening the company’s cash-neutral posture, while elevating chip chief Johny Srouji and posting record R&D, looks like a carefully staged way to prepare Wall Street for bigger spending on AI, silicon, or even acquisitions.
Siri’s next test is coming fast
Before moving on, Alex says Apple still has the basic problem of a bad AI product today, and the next Siri update will be a reality check. MG says Apple could be holding it back for a “big wow” moment, but it could just as easily mean the same ugly integration and edge-case problems that have tripped up Google Assistant, Bard/Gemini, and Alexa’s transition too.
Microsoft and OpenAI are acting like exes trying to stay on good terms
In the second half, Alex and MG unpack the latest Microsoft-OpenAI rewrite: the AGI clause is gone, Microsoft keeps access to OpenAI IP through 2032, and OpenAI can now work with other cloud providers. MG’s read is blunt — the relationship has clearly been tense for a while, and this looks like both sides deciding they’re better off with more distance, even if Satya and Sam still smile for the photos.
OpenAI’s cloud freedom comes with Azure’s fingerprints still on it
MG zeroes in on the key caveat: Microsoft remains OpenAI’s primary cloud partner, and OpenAI products ship first on Azure unless Microsoft cannot or chooses not to support them. So yes, OpenAI can sell on AWS now and probably GCP next, but Azure may still get the “now in theaters” window before everyone else gets the DVD.
Stargate went from moonshot infrastructure plan to messy reality
Finally, they revisit Stargate — the $500 billion infrastructure initiative launched at the White House with Trump, Altman, Masayoshi Son, and Larry Ellison. MG says the big shift is that OpenAI no longer seems to be building and owning a giant proprietary network of data centers; instead, “Stargate” has become a catch-all label for infrastructure secured through partners, with risk increasingly pushed onto Oracle, Nvidia, SoftBank, and neocloud players while OpenAI quietly rewrites the story around what was promised.
Share
Keep Reading
Two Ways to Follow
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.
